A CI/CD pipeline with tests, Docker and blue-green release

Tiffin Trail runs a small food-ordering web service. One developer deploys it by copying files to a single server at midnight and restarting it, hoping nothing breaks. Twice last month the site was down for 20 minutes because a bad deploy went straight to live users, and there was no quick way back.

The team wants deploys to be boring. They want every change tested automatically before it can ship, the application packaged the same way every time, and a release that can be undone in seconds if it misbehaves. They do not have a cluster and do not want one yet; they have two small environments and a container runtime.

You will build the pipeline around a small web service, which you can write or take as a given starter. The pipeline runs on every push: it installs dependencies, runs the tests and a linter, builds a Docker image tagged with the commit, and pushes it. A deploy step then runs the new image alongside the old one, checks it is healthy, switches traffic to it, and keeps the old one ready for a moment. If the health check fails, traffic never switches. Done looks like a green pipeline on every push, an image that runs the same on any machine, a blue-green switch that a reviewer can watch, and a rollback command that returns to the previous image in under a minute. The team signs off when a deliberately broken build is stopped by the pipeline and never reaches the running site.